Some electronic systems include devices (e.g., radio frequency (RF) power amplifiers), which add distortion to a signal during processing. Such devices are referred to as “non-linear” devices when the added distortion is non-linear in nature. Distortion may include, for example, variations in phase differences and/or variations in amplitude differences. For example, one type of non-linearity occurs when the phase between the input and output signals varies with input signal amplitude. This is generally referred to as amplitude-to-phase distortion (am/pm). Another type of non-linearity occurs when the ratio of output signal amplitude to input signal amplitude varies with input signal amplitude. This is generally referred to as amplitude-to-amplitude distortion (am/am). Significant distortion, left uncompensated for, may result in poor signal quality.
Digital pre-distortion (DPD) systems have been used in the past to compensate for the intrinsic distortion characteristics of non-linear devices. A traditional DPD system determines an error signal, which reflects differences between an input signal and a feedback signal from the system output. “Insertion phase” is a term used to describe a quantity representing an average difference in phase between the input signal and the feedback signal, or equivalently, the average phase of the error signal. The error signal is used to determine a complementary distortion or inverse gain signal. The inverse gain signal, which inversely reflects the error signal, is combined with the input signal to produce a “pre-distorted” signal. In many cases, this process results in effective cancellation of the distortion (i.e., the non-linearities) produced within the system, and a more linear output signal may result.
Lookup tables have been used to store inverse gain values. In some traditional DPD systems, an initialization and training process is performed prior to transmission of actual input data, in order to determine the inverse gain values within the lookup table. Using this process, a training sequence is applied to the system, during which time the inverse gain values are adapted to reflect the system's distortion characteristics. After completion of the training process, actual input data is then combined with the adapted inverse gain values, and the combined signal is transmitted. In other traditional DPD systems, a training process is not performed. Instead, the inverse gain values in the lookup table are set to pre-defined, “blind” settings. The blind settings do not reflect the actual, then-current distortion characteristics, but instead reflect a pre-defined estimate of the distortion characteristics. At the onset of a transmission burst, the blind settings are applied to the actual input data to attempt to compensate for the distortion. Eventually, the inverse gain values may adapt to more accurately reflect the actual distortion characteristics.
The prior DPD systems described above suffer from several drawbacks. In particular, using prior DPD systems, when an error signal reflects a significant insertion phase (e.g., at the onset of a transmission), convergence to an acceptable pre-distorted signal may take a long time to occur. In some situations, convergence may not occur at all. In addition, DPD systems that use training processes may add considerable complexity, hardware, and cost to transmitter systems. In addition, timing issues (e.g., issues relating to the burst/slot/frame structure) may preclude the use of training processes. Although, DPD systems that use “blind” settings of initial lookup table values may overcome some of these disadvantages, they suffer from other disadvantages. For example, the linearization performance of these DPD systems are strongly coupled to the choice of initial inverse gain values in the lookup table. In environments in which the distortion characteristics are not well known and/or change considerably with environmental conditions and/or transmitter operation, poor initial performance may result from inaccurate initial inverse gain values. In some cases, when the initial inverse gain values are substantially inaccurate, the system may not be able to converge at all. For at least the above reasons, a need exists for DPD systems and methods that provide relatively fast convergence to an acceptable pre-distorted signal, even in the face of relatively large insertion phases.